Reproducing results from the paper

[Wayne2Wang]

Dear authors,

Thank you for publishing your code for the great work! The code is easily readable and well structured, but I do have a few minor questions about it:

1. It was mentioned in the paper that you used a Lognormal prior with an std of 0.5 in the SinkhornKnopp algorithm. However, it seems like the code uses 0.1, which was set to the default value and never changed in the config files. I wonder which value you would recommend using to reproduce the results.
2. As was pointed out by a previous issue, the it180k_bs16.py file is missing inside the configs.train.fully_supervised folder. I was able to write my own version of it according to the descriptions in the paper, but I would appreciate it if an official version could be provided.

Many thanks for considering my request.